Re: 0005-Part-5.-Add-x86-CET-documentation
On 09/27/2017 09:17 AM, Tsimbalist, Igor V wrote: Updated version #3. diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi index e52a1ea..accba40 100644 --- a/gcc/doc/extend.texi +++ b/gcc/doc/extend.texi @@ -5655,6 +5655,14 @@ compiled with the @option{-fcf-protection=branch} option. The compiler assumes that the function's address is a valid target for a control-flow transfer. +@emph{x86 implementation:} when @option{-fcf-protection} option is +specified the compiler inserts an @code{endbr} instruction at function's +prologue if the function's type does not have the @code{nocf_check} +attribute and addresses to which indirect control-flow transfer can +happen. The instruction triggers the HW check if a control-flow +transfer to the address where @code{endbr} instruction was inserted +is valid. + I think the consensus among Joseph, Jeff, and I is that this doesn't belong in the GCC manual at all, but in the ABI documentation. So please delete the implementation note. @@ -5662,7 +5670,9 @@ not be instrumented when compiled with the that the function's address from the pointer is a valid target for a control-flow transfer. A direct function call through a function name is assumed to be a safe call thus direct calls are not -instrumented by the compiler. +instrumented by the compiler. For @emph{x86 implementation} the +compiler inserts a @code{notrack} prefix before an indirect call +instruction. Ditto with this implementation note. diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi index c4faa23..189130b 100644 --- a/gcc/doc/invoke.texi +++ b/gcc/doc/invoke.texi @@ -1203,6 +1203,7 @@ See RS/6000 and PowerPC Options. -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop @gol -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx @gol -mmwaitx -mclzero -mpku -mthreads @gol +-mcet -mibt -mshstk @gol -mms-bitfields -mno-align-stringops -minline-all-stringops @gol -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol @@ -11374,6 +11375,14 @@ You can also use the @code{nocf_check} attribute to identify which functions and calls should be skipped from instrumentation (@pxref{Function Attributes}). +Currently the x86 GNU/Linux target provides an implementation based +on Intel Control-flow Enforcement Technology (CET). Instrumentation +for x86 is controlled by target-specific options @option{-mcet}, +@option{-mibt} and @option{-mshstk} (@pxref{x86 Options}). This part is OK. +The compiler also provides a number of built-in functions for +fine-grained control in a CET-based application. +See @xref{x86 Built-in Functions}, for more information. I think these builtins emit instructions in the CET extension explicitly and don't affect the GCC's code generation for the -fcf-protection option. So please move this to the discussion of -mcet in the x86 options section instead @@ -25779,6 +25792,11 @@ supported architecture, using the appropriate flags. In particular, the file containing the CPU detection code should be compiled without these options. +The @option{-mcet} option turns on the @option{-mibt} and @option{-mshstk} +options. The @option{-mibt} option enables indirect branch tracking support +and the @option{-mshstk} option enables shadow stack support from +Intel Control-flow Enforcement Technology (CET). + ...here. The patch is OK with those changes. -Sandra
Re: 0005-Part-5.-Add-x86-CET-documentation
On 09/27/2017 09:17 AM, Tsimbalist, Igor V wrote: >> > > 0005-Add-x86-CET-documentation.patch > > > From dda22b06a3a5bde9b0dc57585d878db520769510 Mon Sep 17 00:00:00 2001 > From: Igor Tsimbalist> Date: Tue, 4 Jul 2017 13:55:03 +0300 > Subject: [PATCH 5/6] Add x86 CET documentation. > > gcc/ > * doc/extend.texi: Add x86 specific to 'nocf_check' attribute. > List CET intrinsics. > * doc/invoke.texi: Add -mcet, -mibt, -mshstk options. Add x86 > specific to -fcf-protection option. Once Sandra has OK'd the doc patches, they should be considered OK for the trunk. jeff
Re: 0005-Part-5.-Add-x86-CET-documentation
On 09/27/2017 11:01 AM, Joseph Myers wrote: > On Wed, 27 Sep 2017, Florian Weimer wrote: > >> This is part of the ABI GCC implements, so it has to be documented somewhere, >> and not just as part of the GCC source code. >> >> CET is not properly described in the ABI supplement and I don't think this >> will change, so detailed documentation in the GCC manual is very much >> desirable. > > Isn't this a matter to take up further in the thread HJ started on the ABI > mailing lists, or a new such thread (possibly e.g. sending pull requests > that build further on his wording, or propose alternative wording, to > clarify them things left unclear there, with a goal of getting it clearly > defined in the master sources for x86_64 and x86)? Clearly the best > result would be proper documentation in the ABI and the GCC manual > cross-referencing the relevant ABI documents. The documentation should be AFAICT independent of the compiler in use -- ie, gcc, llvm and icc all should agree on where/when these new instructions should be inserted. Which argues that the documentation belongs in the ABI docs, not the GCC docs. *users* aren't really going to care about these kinds of details. So I think the summary is that I agree with Joseph on this. Let's push it into the ABI docs. HJ can and should play a central role in making that happen. jeff
Re: 0005-Part-5.-Add-x86-CET-documentation
On Wed, 27 Sep 2017, Florian Weimer wrote: > This is part of the ABI GCC implements, so it has to be documented somewhere, > and not just as part of the GCC source code. > > CET is not properly described in the ABI supplement and I don't think this > will change, so detailed documentation in the GCC manual is very much > desirable. Isn't this a matter to take up further in the thread HJ started on the ABI mailing lists, or a new such thread (possibly e.g. sending pull requests that build further on his wording, or propose alternative wording, to clarify them things left unclear there, with a goal of getting it clearly defined in the master sources for x86_64 and x86)? Clearly the best result would be proper documentation in the ABI and the GCC manual cross-referencing the relevant ABI documents. -- Joseph S. Myers jos...@codesourcery.com
Re: 0005-Part-5.-Add-x86-CET-documentation
On 09/27/2017 02:52 AM, Florian Weimer wrote: On 09/27/2017 05:40 AM, Sandra Loosemore wrote: +@emph{x86 implementation:} when @option{-fcf-protection} option is +specified the compiler inserts an ENDBR instruction at function's +prologue if the function's type does not have the @code{nocf_check} +attribute and addresses to which indirect control-flow transfer can +happen. The instruction triggers the HW check if a control-flow +transfer to the address of ENDBR instruction is valid. Implementation details like this should be comments in the code, not included in the user-facing documentation. This is part of the ABI GCC implements, so it has to be documented somewhere, and not just as part of the GCC source code. CET is not properly described in the ABI supplement and I don't think this will change, so detailed documentation in the GCC manual is very much desirable. Not if you're a documentation maintainer. :-( Generally speaking, user-facing manuals like the GCC manual should document user-visible GCC features, not internal implementation details. Especially the target-independent parts of the manual are not the right place to discuss target-specific code generation patterns or conventions that should be in the ABI supplement or some other non-GCC documentation. I don't have so much objection to expanding the discussion of the target-specific -mcet option in the x86 options section, as long as the documentation is there because it helps people *use* the feature and not to explain things that are only interesting to compiler implementors. -Sandra
RE: 0005-Part-5.-Add-x86-CET-documentation
Updated version #3. > -Original Message- > From: Sandra Loosemore [mailto:san...@codesourcery.com] > Sent: Wednesday, September 27, 2017 5:41 AM > To: Tsimbalist, Igor V <igor.v.tsimbal...@intel.com>; Uros Bizjak > <ubiz...@gmail.com> > Cc: gcc-patches@gcc.gnu.org > Subject: Re: 0005-Part-5.-Add-x86-CET-documentation > > On 09/26/2017 07:47 AM, Tsimbalist, Igor V wrote: > > Here is a new version of the patch. > > > > diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi index > > a374890..a900ed1 100644 > > --- a/gcc/doc/extend.texi > > +++ b/gcc/doc/extend.texi > > @@ -5655,6 +5655,13 @@ compiled with the > > @option{-fcf-protection=branch} option. The compiler assumes that > > the function's address is a valid target for a control-flow transfer. > > > > +@emph{x86 implementation:} when @option{-fcf-protection} option is > > +specified the compiler inserts an ENDBR instruction at function's > > +prologue if the function's type does not have the @code{nocf_check} > > +attribute and addresses to which indirect control-flow transfer can > > +happen. The instruction triggers the HW check if a control-flow > > +transfer to the address of ENDBR instruction is valid. > > Implementation details like this should be comments in the code, not > included in the user-facing documentation. > > > @@ -5662,7 +5669,8 @@ not be instrumented when compiled with the > that > > the function's address from the pointer is a valid target for a > > control-flow transfer. A direct function call through a function > > name is assumed to be a safe call thus direct calls are not > > -instrumented by the compiler. > > +instrumented by the compiler. For @emph{x86 implementation} the > > +compiler inserts a NOTRACK prefix before an indirect call instruction. > > Likewise here. For this comment and above could you please let me know what is the right place To move the description? Also I enclosed ENDBR and NOTRACK in @code{} and wrote it in lower case. > > @@ -21217,6 +21225,25 @@ void __builtin_ia32_wrpkru (unsigned int) > > unsigned int __builtin_ia32_rdpkru () @end smallexample > > > > +The following built-in functions are available when @option{-mcet} is > used. > > +They are used to support Intel Control-flow Enforcment Technology (CET). > > +Each built-in function generate a machine instruction that is part of > > +the > > s/generate a/generates the/ Fixed. > > @@ -11378,6 +11379,20 @@ You can also use the @code{nocf_check} > > attribute to identify which functions and calls should be skipped > > from instrumentation (@pxref{Function Attributes}). > > > > +Currently x86 GNU/Linux target provides an implementation based on > > s/x86/the x86/ Fixed. > > +Intel Control-flow Enforcement Technology (CET), thus @option{-mcet} > > s/@option/the @option/ Fixed. > > +option is required to enable this feature. > > I think you should put a cross-reference to the x86 options node here, and > move all the following x86-specific discussion to that section. Put cross-reference. > > In order to get an > > +application to be CET compatible the x86 implementation requires all > > +object files have to be compiled with @option{-fcf-protection} option > > +and all linked in libraries have to be CET compatible. > > I'm having difficulty parsing this. What does "CET compatible" mean? > Is this an ABI compatibility issue, so that all objects linked into the > executable > have to be compiled with the (same?) @option{-fcf-protection} option if any > of them do? Or do you just lose checking on code in uninstrumented > objects? I re-wrote the paragraph and removed "compatibility topic". > > +Instrumentation for x86 is controlled by target specific options > > hyphenate target-specific here Fixed. > > +@option{-mcet}, @option{-mibt} and @option{-mshstk}. The compiler > > +also provides a number of built-in functions for fine-grained control > > +of CET-based implementation. See @xref{x86 Built-in Functions}, for > > +more information. > > + > > @item -fstack-protector > > @opindex fstack-protector > > Emit extra code to check for buffer overflows, such as stack smashing > > @@ -25755,15 +25770,19 @@ preferred alignment to @option{- > mpreferred-stack-boundary=2}. > > @need 200 > > @itemx -mclzero > > @opindex mclzero > > +@need 200 > > @itemx -mpku > > @opindex mpku > > +@need 200 > > +@itemx -mcet > > +@opindex mcet > > These switches enable the use of instructions in the MMX, SSE, SSE2, > >
RE: 0005-Part-5.-Add-x86-CET-documentation
> -Original Message- > From: Florian Weimer [mailto:fwei...@redhat.com] > Sent: Wednesday, September 27, 2017 10:52 AM > To: Sandra Loosemore <san...@codesourcery.com>; Tsimbalist, Igor V > <igor.v.tsimbal...@intel.com>; Uros Bizjak <ubiz...@gmail.com> > Cc: gcc-patches@gcc.gnu.org > Subject: Re: 0005-Part-5.-Add-x86-CET-documentation > > On 09/27/2017 05:40 AM, Sandra Loosemore wrote: > >> > >> +@emph{x86 implementation:} when @option{-fcf-protection} option is > >> +specified the compiler inserts an ENDBR instruction at function's > >> +prologue if the function's type does not have the @code{nocf_check} > >> +attribute and addresses to which indirect control-flow transfer can > >> +happen. The instruction triggers the HW check if a control-flow > >> +transfer to the address of ENDBR instruction is valid. > > > > Implementation details like this should be comments in the code, not > > included in the user-facing documentation. > > This is part of the ABI GCC implements, so it has to be documented > somewhere, and not just as part of the GCC source code. A question for both Sandra and Florian - What is your suggestion where the text should go? > CET is not properly described in the ABI supplement and I don't think this > will > change, so detailed documentation in the GCC manual is very much > desirable. > > That being said, the implementation notes above need some clarification. > It's not clear to me what the conditions are under which the ENDBR > instruction is emitted (and we probably should use @code{endbr} in the > manual), what it is trying to achieve, and how the x86 calling convention > changes. I assume it is somehow related to what we call internally “the > suffix We are diving into implementation details but it's simple enough. - endbr is generated for every function, which does not have nocf_check attribute. Optimization can be done later to exclude functions, whose address was not taken. - there is no change in calling convention Thanks, Igor > problem”: without control flow integrity, an attacker might skip over > precondition/hardening checks, directly to the critical changes we want to > protect, executing only the suffix of a function (hence the name). > > Thanks, > Florian
Re: 0005-Part-5.-Add-x86-CET-documentation
On 09/27/2017 05:40 AM, Sandra Loosemore wrote: +@emph{x86 implementation:} when @option{-fcf-protection} option is +specified the compiler inserts an ENDBR instruction at function's +prologue if the function's type does not have the @code{nocf_check} +attribute and addresses to which indirect control-flow transfer can +happen. The instruction triggers the HW check if a control-flow +transfer to the address of ENDBR instruction is valid. Implementation details like this should be comments in the code, not included in the user-facing documentation. This is part of the ABI GCC implements, so it has to be documented somewhere, and not just as part of the GCC source code. CET is not properly described in the ABI supplement and I don't think this will change, so detailed documentation in the GCC manual is very much desirable. That being said, the implementation notes above need some clarification. It's not clear to me what the conditions are under which the ENDBR instruction is emitted (and we probably should use @code{endbr} in the manual), what it is trying to achieve, and how the x86 calling convention changes. I assume it is somehow related to what we call internally “the suffix problem”: without control flow integrity, an attacker might skip over precondition/hardening checks, directly to the critical changes we want to protect, executing only the suffix of a function (hence the name). Thanks, Florian
Re: 0005-Part-5.-Add-x86-CET-documentation
On 09/26/2017 07:47 AM, Tsimbalist, Igor V wrote: Here is a new version of the patch. diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi index a374890..a900ed1 100644 --- a/gcc/doc/extend.texi +++ b/gcc/doc/extend.texi @@ -5655,6 +5655,13 @@ compiled with the @option{-fcf-protection=branch} option. The compiler assumes that the function's address is a valid target for a control-flow transfer. +@emph{x86 implementation:} when @option{-fcf-protection} option is +specified the compiler inserts an ENDBR instruction at function's +prologue if the function's type does not have the @code{nocf_check} +attribute and addresses to which indirect control-flow transfer can +happen. The instruction triggers the HW check if a control-flow +transfer to the address of ENDBR instruction is valid. Implementation details like this should be comments in the code, not included in the user-facing documentation. @@ -5662,7 +5669,8 @@ not be instrumented when compiled with the that the function's address from the pointer is a valid target for a control-flow transfer. A direct function call through a function name is assumed to be a safe call thus direct calls are not -instrumented by the compiler. +instrumented by the compiler. For @emph{x86 implementation} the +compiler inserts a NOTRACK prefix before an indirect call instruction. Likewise here. @@ -21217,6 +21225,25 @@ void __builtin_ia32_wrpkru (unsigned int) unsigned int __builtin_ia32_rdpkru () @end smallexample +The following built-in functions are available when @option{-mcet} is used. +They are used to support Intel Control-flow Enforcment Technology (CET). +Each built-in function generate a machine instruction that is part of the s/generate a/generates the/ @@ -11378,6 +11379,20 @@ You can also use the @code{nocf_check} attribute to identify which functions and calls should be skipped from instrumentation (@pxref{Function Attributes}). +Currently x86 GNU/Linux target provides an implementation based on s/x86/the x86/ +Intel Control-flow Enforcement Technology (CET), thus @option{-mcet} s/@option/the @option/ +option is required to enable this feature. I think you should put a cross-reference to the x86 options node here, and move all the following x86-specific discussion to that section. In order to get an +application to be CET compatible the x86 implementation requires +all object files have to be compiled with +@option{-fcf-protection} option and all linked in libraries have +to be CET compatible. I'm having difficulty parsing this. What does "CET compatible" mean? Is this an ABI compatibility issue, so that all objects linked into the executable have to be compiled with the (same?) @option{-fcf-protection} option if any of them do? Or do you just lose checking on code in uninstrumented objects? +Instrumentation for x86 is controlled by target specific options hyphenate target-specific here +@option{-mcet}, @option{-mibt} and @option{-mshstk}. The compiler +also provides a number of built-in functions for fine-grained control +of CET-based implementation. See @xref{x86 Built-in Functions}, +for more information. + @item -fstack-protector @opindex fstack-protector Emit extra code to check for buffer overflows, such as stack smashing @@ -25755,15 +25770,19 @@ preferred alignment to @option{-mpreferred-stack-boundary=2}. @need 200 @itemx -mclzero @opindex mclzero +@need 200 @itemx -mpku @opindex mpku +@need 200 +@itemx -mcet +@opindex mcet These switches enable the use of instructions in the MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD, SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR, -XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, 3DNow!@: or enhanced 3DNow!@: -extended instruction sets. Each has a corresponding @option{-mno-} option -to disable use of these instructions. +XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU, IBT, SHSTK, +3DNow!@: or enhanced 3DNow!@: extended instruction sets. Each has a +corresponding @option{-mno-} option to disable use of these instructions. These extensions are also available as built-in functions: see @ref{x86 Built-in Functions}, for details of the functions enabled and @@ -25783,6 +25802,11 @@ supported architecture, using the appropriate flags. In particular, the file containing the CPU detection code should be compiled without these options. +The @option{-mcet} option turns on @option{-mibt} and @option{-mshstk} s/turns on/turns on the/ +options. @option{-mibt} option enables idirect branch tracking support s/@option/The @option/ s/idirect/indirect/ +and @option{-mshstk} option enables shadow stack support from s/@option/the @option/ +Intel Control-flow Enforcement Technology (CET). + @item -mdump-tune-features @opindex mdump-tune-features This option instructs GCC to dump the names of the x86
RE: 0005-Part-5.-Add-x86-CET-documentation
Here is a new version of the patch. Igor > -Original Message- > From: Sandra Loosemore [mailto:san...@codesourcery.com] > Sent: Monday, September 25, 2017 5:43 AM > To: Uros Bizjak <ubiz...@gmail.com>; Tsimbalist, Igor V > <igor.v.tsimbal...@intel.com> > Cc: gcc-patches@gcc.gnu.org > Subject: Re: 0005-Part-5.-Add-x86-CET-documentation > > On 09/20/2017 08:13 AM, Uros Bizjak wrote: > > On Wed, Sep 20, 2017 at 11:20 AM, Tsimbalist, Igor V > > <igor.v.tsimbal...@intel.com> wrote: > >> Uros, could you please review this patch as it's a part of x86 specific > changes you have reviewed already. > > > > Please proofread and spell-check the text. There are grammatical errors, > e.g.: > > [snip] > > If/when there is a newer version of this patch, cc it to me and I will also > make > a pass through it. > > -Sandra 0005-Part-5.-Add-x86-CET-documentation.patch Description: 0005-Part-5.-Add-x86-CET-documentation.patch
Re: 0005-Part-5.-Add-x86-CET-documentation
On 09/20/2017 08:13 AM, Uros Bizjak wrote: On Wed, Sep 20, 2017 at 11:20 AM, Tsimbalist, Igor Vwrote: Uros, could you please review this patch as it's a part of x86 specific changes you have reviewed already. Please proofread and spell-check the text. There are grammatical errors, e.g.: [snip] If/when there is a newer version of this patch, cc it to me and I will also make a pass through it. -Sandra
Re: 0005-Part-5.-Add-x86-CET-documentation
On Wed, Sep 20, 2017 at 11:20 AM, Tsimbalist, Igor Vwrote: > Uros, could you please review this patch as it's a part of x86 specific > changes you have reviewed already. Please proofread and spell-check the text. There are grammatical errors, e.g.: +When compiled with the @option{-finstrument-control-flow} option +the ENDBR instruction also inserted at addresses to which indirect and typos, such as: +control-flow trasfer can happen. English is not my tongue, but these two popped out immediately. Other than that, -a program security by checking a target address of control-flow +a program security by checking target addresses of control-flow Spurious change. +@option{mshstk} options. Each of these options enables coresponding Missing dash, and a typo. +This option instructs the compiler to use only indirect branch tracking +(indirect calls and jumps) support from x86 Control-flow Enforcment Typo. +Technology (CET) implementation, see also @code{-mcet} option. The +option has effect only if @code{-finstrument-control-flow} option is +specified. The option @code{-mibt} is on by default when @code{-mcet} +option is specified. Please use @option{} for options. +This option instructs compiler to use only shadow stack (return address +tracking) support from x86 Control-flow Enforcment Technology (CET) +implementation, see also @code{-mcet} option. The option has effect +only if @code{-finstrument-control-flow} option is specified. The +option @code{-mshstk} is on by default when @code{-mcet} option is +specified. Also in the above text. Uros.
RE: 0005-Part-5.-Add-x86-CET-documentation
Uros, could you please review this patch as it's a part of x86 specific changes you have reviewed already. Thanks, Igor > -Original Message- > From: Tsimbalist, Igor V > Sent: Tuesday, August 1, 2017 10:57 AM > To: 'gcc-patches@gcc.gnu.org'> Cc: Tsimbalist, Igor V > Subject: 0005-Part-5.-Add-x86-CET-documentation > > Part#5. Add x86 CET documentation.